Approximating Feynman integrals using complete monotonicity and Stieltjes properties

Prashanth Raman · 2026

Read Approximating Feynman integrals using complete monotonicity and Stieltjes properties by Prashanth Raman (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-27101.txt.

Introduction

Raman presents two fresh ways to compute Feynman integrals, the very hard integrals behind particle physics predictions. Both exploit positivity, a shared feature of these integrals in the Euclidean region, where lengths stay real and positive. Complete monotonicity means every slope of the function keeps a fixed sign, a strong shape constraint. The first approach, a complete monotonicity bootstrap, rebuilds integrals from their governing rules with no boundary data supplied. Rigorous bounds come out of that reconstruction, rather than mere guesses.

A stronger result powers the second approach: within a stated parameter range, these integrals are Stieltjes functions, a narrower positive class. That class suits Padé approximants, ratio estimates with proven convergence across the cut complex plane. Fast computing and analytic continuation, extending values into new regions, both follow. Examples include the simple massive bubble integral and the twenty-loop banana integral.

Catalog record

CREATORPrashanth Raman
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE14 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.27101

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Approximating Feynman integrals using complete monotonicity and Stieltjes properties Prashanth Raman Abstract We present two novel approaches for the numerical evaluation of Feynman integrals based on their universal analytic properties related to positivity, namely complete monotonicity (CM) and Stieltjes properties. Building on recent results, we exploit the fact that scalar Feynman integrals in the Euclidean region are completely monotonic functions, meaning that all their derivatives have a fixed sign. Building on this observation, the CM bootstrap allows one to reconstruct integrals from differential equations without explicit boundary data, yielding rigorous bounds. The second method is based on a refinement of CM. We prove that Feynman integrals, within a certain range of parameters, are not only CM but in fact Stieltjes functions. This enables the use of Padé approximants with provable convergence properties in the cut complex plane, providing an efficient method for analytic continuation and fast numerical evaluation. We illustrate the method with simple examples such as the massive bubble integral and discuss applications to multi-loop integrals, including the 20-loop banana integral. Finally, we comment on a number of extensions of these novel avenues for computing Feynman integrals. This talk is based on [ 6 ] . [a,b]Sara Ditsch 1 Introduction The numerical evaluation of multi-loop Feynman integrals is a central problem in perturbative quantum field theory. A variety of methods have been developed, including sector decomposition, differential equation techniques, auxiliary mass flow, and Monte Carlo integration, see for example [ 4 , 8 , 15 , 5 , 10 , 1 , 14 , 12 , 11 ] . Despite these advances, efficient and robust evaluation of multi-loop integrals across kinematic regions remains challenging. In this contribution, we present two novel methods for approximating Feynman integrals based on universal analytic properties related to positivity. The first is based on the observation that scalar Feynman integrals in the Euclidean region satisfy complete monotonicity (CM), which was shown in [ 9 ] . Combining this with the differential equations satisfied by a set of integrals, yields powerful constraints on the space of possible solutions and can uniquely determine the integrals. We extend this idea, by proving that, under suitable conditions, Feynman integrals belong to the class of Stieltjes functions , which provides an even stronger constraint. These properties can be leveraged to obtain Padé approximants with rigorous error bounds. 2 Bootstrapping Feynman integrals from complete monotonicity 2.1 Complete monotonicity and Feynman integrals A function f:R\to\mathbb{R} is called completely monotonic in the region R if it satisfies (cf. ref. [ 16 ] ) (-1)^{n}\frac{d^{n}}{dx^{n}}f(x)\geq 0\,,\quad\forall\,n\in\mathbb{N}_{0},\;\forall\,x\in R\,. (1) It has recently been shown in [ 9 ] that scalar Feynman integrals in the Euclidean region are completely monotonic functions of appropriate kinematic variables. This property encodes non-trivial positivity constraints that can be used to constrain the function. 2.2 The CM bootstrap The CM bootstrap combines differential equations for Feynman integrals with positivity constraints from complete monotonicity. Feynman integrals satisfy systems of differential equations of the form \frac{d}{dx}{\bf f}(x)=A(x){\bf f}(x)\,, (2) where {\bf f}(x) denotes a vector of master integrals. Traditionally, solving such systems requires boundary conditions. In the CM bootstrap approach, boundary conditions are replaced by positivity constraints. The differential equation allows rewriting the CM property as a matrix equation for scalar Feynman integrals of the form \displaystyle Q_{n}(x){\bf f}(x)\geq 0\,,\quad\forall\,n\in\mathbb{N}_{0}\,, (3) where higher derivatives are obtained from a matrix recursively defined by \displaystyle Q_{0}=\mathbf{1}\,,\quad Q_{1}(x)=-A\,,\quad Q_{n}(x)=-\partial_{x}Q_{n-1}(x)+Q_{n-1}(x)Q_{1}(x)\,. (4) Hence we obtain linear inequalities satisfied by the master integral. Imposing them, leads to rigorous bounds on the integrals in the Euclidean region. In favorable cases, one obtains two-sided bounds with rapid convergence. 2.3 Pedagogical example: The massive bubble integral Let us illustrate this on the example of the massive one-loop bubble integral \displaystyle f(x)=\int\frac{d^{2}k}{i\pi}\frac{1}{(-k^{2}+m^{2})[-(k+p)^{2}+m^{2}]}\,. (5) The corresponding integral family has two master integrals: The bubble and the tadpole integral. For this family, we obtain the differential equation \displaystyle\partial_{x}{\bf f}=A_{x}\,{\bf f}\,,\quad{\rm with}\quad A_{x}=-\begin{pmatrix}0&0\\ -\frac{2}{(4+x)x}&\tfrac{2+x}{(4+x)x}\end{pmatrix}\,, (6) where we set m^{2}=1 and -p^{2}=x . Scale invariance allows fixing the value tadpole. Now, one can impose the CM bootstrap to obtain numerical bounds on the value of the bubble integral in the Euclidean region. The results for n_{0}=5 are shown in Fig. 1 . There exist three regions with qualitatively different behavior. In the first region, where x\in(-4,-2) , we get only a loose lower bound. For x\in(-2,0) , we obtain a two-sided bound, which tightly constrains the value of the function. Note, that the convergence depends strongly on the phase space point. For x>0 , we get only a lower bound. Interestingly, that bound is very close to the value of the function. Figure 1: Constraints obtained from the CM bootstrap for the massive bubble integral. The dashed yellow and dashed blue lines correspond to upper and lower bounds, respectively. The white regions and gray regions indicate allowed and forbidden regions, respectively. The solid black line corresponds to the exact function value. 2.4 Application to multi-loop banana integrals We also show, how the CM bootstrap can be applied to banana integrals up to four loops. The resulting bounds are qualitatively similar to those obtained for the bubble integral. Despite the analytic complexity of the function space due to Calabi-Yau-Manifolds, the evaluation time is similar to that of the bubble and only seems to scale with the number of master integrals. This is related to the fact, that the input remains the differential equation matrix, which has purely rational entries. We also compare the evaluation time against AMFlow [ 11 ] . As for the bubble integral, the convergence of our bootstrap depends strongly on the phase space point. While it is rather slow on one side of the interval with a two-sided bound, it can easily outperform AMFlow on the other side. 3 Padé approximants for Stieltjes Feynman integrals We now show, how to extend these ideas to obtain values in the full cut complex plane using Padé approximants. 3.1 Stieltjes functions and Padé approximants We first introduce Stieltjes functions, that form a subclass of completely monotonic functions and are defined by admitting an integral representation of the form f(z)=\int_{0}^{\tfrac{1}{R}}\frac{\rho(u)}{1+uz}\,du\,,\quad{\rm{with}}\quad\rho(u)\geq 0\;\forall\;u\in(0,1/R)\,. (7) A key advantage of Stieltjes functions is that their Padé approximants converge rapidly in the complex plane. In particular, P^{N-1}_{N}(x;x_{0}) and P^{N}_{N}(x;x_{0}) provide upper and lower bounds, respectively, to the function for values on the real axis, that converge towards the function value. More generally, they converge towards the function value with rigorous error bounds in the cut complex plane, see [ 2 , 3 ] . 3.2 Feynman integrals are Stieltjes A key result is that certain scalar Feynman integrals satisfy the defining properties of Stieltjes functions. In Feynman parametrization, Feynman integrals take the schematic form \displaystyle I(\{x_{i}\})\sim\int_{0}^{\infty}\frac{\prod_{i}d\alpha_{i}~\alpha_{i}^{\nu_{i}-1}}{\rm{GL}(1)}\frac{U^{q}}{F^{e}}\,, (8) with U being a non-negative homogeneous polynomial in the integration variables \alpha_{i} , we showed that these integrals are Stieltjes, provided that the exponent of the second Symanzik polynomial satisfies 0<\sum_{i}\nu_{i}-LD/2\leq 1 , where L is the loop order, D is the space-time dimension, and \nu_{i} are the propagator powers. For any family of Feynman integrals it is usually possible to chose propagator powers that satisfy this constraint. A second condition demands that the Euclidean region exists, i.e. there is a choice of variables for

Read the complete text